Spatial Multivariate Trees

Peruzzi M, Dunson DB (2022) Spatial Multivariate Trees for Big Data Bayesian Regression. Journal of Machine Learning Research 23(17):1–40.

When multivariate spatial outcomes are very misaligned and some are very sparse, we show that recursive treed Directed Acyclic Graphs (DAGs) can be used to efficiently model multivariate spatial dependence and build scalable sampling algorithms.

Data application: Multivariate regression model on MODIS satellite data and GHCN climate station data over the continental United States.